Catálogo de publicaciones - libros

This paper proposes a new support vector machine (SVM) with a robust loss function for data mining. Its dual optimal formation is also constructed. A gradient based algorithm is designed for fast and simple implementation of the new support vector machine. At the same time it analyzes algorithm’s convergence condition and gives a formula to select learning step size. Numerical simulation results show that the new support vector machine performs significantly better than a standard support vector machine.

This paper develops an infinite polynomial kernel for support vector machines. We also propose a mapping from an original data space into the high dimensional feature space on which the inner product is defined by the infinite polynomial kernel . Via this mapping, any two finite sets of data in the original space will become linearly separable in the feature space. Numerical experiments indicate that the proposed infinite polynomial kernel possesses some properties and performance better than the existing finite polynomial kernels.

QOSPF (Quality of Service Open Shortest Path First) based on QoS routing has been recognized as a missing piece in the evolution of QoS-based service offerings in the Internet. A data mining method for QoS routing based on rough set theory has been presented in this paper. The information system about the link is created from the subnet, and the method of rough set can mine the best route from enormous irregular link QoS data and can classify the link with the link-status data. An instance applying to the theory is presented, which verifies the feasibility that the most excellent attribute set is mined by rough set theory for compatible data table.

Data mining has become of great importance owing to ever-increasing amounts of data collected by large organizations. This paper propose an data mining algorithm called Ant-Miner(I),which is based on an improvement of Ant Colony System(ACS) algorithm. Experimental results show that Ant-Miner(I) has a higher predictive accuracy and much smaller rule list than the original Ant-Miner algorithm.

A precondition of existing ensemble-based distributed data mining techniques is the assumption that contributing data are identically and independently distributed. However, this assumption is not valid in many virtual organization contexts because contextual heterogeneity exists. Focusing on regression tasks, this paper proposes a context-based meta-learning technique for horizontally partitioned data with contextual heterogeneity. The predictive performance of our new approach and the state of the art techniques are evaluated and compared on both simulated and real-world data sets.

Customer Churn Prediction is an increasingly pressing issue in today’s ever-competitive commercial arena. Although there are several researches in churn prediction, but the accuracy rate, which is very important to business, is not high enough. Recently, Support Vector Machines (SVMs), based on statistical learning theory, are gaining applications in the areas of data mining, machine learning, computer vision and pattern recognition because of high accuracy and good generalization capability. But there has no report about using SVM to Customer Churn Prediction. According to churn data set characteristic, the number of negative examples is very small, we introduce an improved one-class SVM. And we have tested our method on the wireless industry customer churn data set. Our method has been shown to perform very well compared with other traditional methods, ANN, Decision Tree, and Naïve Bays.

Feature selection is one of important and frequently used techniques in data preprocessing. It can improve the efficiency and the effectiveness of data mining by reducing the dimensions of feature space and removing the irrelevant and redundant information. Feature selection can be viewed as a global optimization problem of finding a minimum set of M relevant features that describes the dataset as well as the original N attributes. In this paper, we apply the adaptive partitioned random search strategy into our feature selection algorithm. Under this search strategy, the partition structure and evaluation function is proposed for feature selection problem. This algorithm ensures the global optimal solution in theory and avoids complete randomness in search direction. The good property of our algorithm is shown through the theoretical analysis.

Execution cost of batched data mining queries can be reduced by integrating their I/O steps. Due to memory limitations, not all data mining queries in a batch can be executed together. In this paper we introduce a heuristic algorithm called CCFull,which suboptimally schedules the data mining queries into a number of execution phases. The algorithm significantly outperforms the optimal approach while providing a very good accuracy.

Masquerade attacks may be one of the most serious attacks in computer security context. To avoid being detected, masqueraders sometimes insert some common commands such as “ls” into their command sequences intentionally for concealing their actual purpose. This causes the masquerade attacks difficult to be detected. We refer to these command sequences mixed with confusable commands as gap-insensitive. To eliminate the effects on the insertion, we present a string kernel called gap-insensitive kernel without regard to the gaps in the command sequences, and use it to detect masquerade attacks. We test it and other kernels on the dataset from keyboard commands on a UNIX platform. We find that many users’ attacks against other users can be easily detected by our gap-insensitive kernel, which means that the command sequences of these attackers are gap-insensitive. The results reveal that gap-insensitive kernel can determine gap-insensitivity in command sequences, and efface the gaps in the sequences.

In order to overcome the weaknesses of Regularization Method for linear ill-posed problem, the authors suggest a new method named Iteration Method by Rectifying Eigenvalue (IMRE) in this paper. Firstly, the rigorous theoretical proofs that IMRE can achieve convergent and unbiased solution are given. Then an effective method called L-Curve method is introduced to determine parameter in IMRE. Thirdly, a computing program is designed. Finally an example is given to testify the advantages of IMRE by the above program.